LangPlayground/Python/similar-images-classification/main.py

import os
import sys
import numpy as np
from tensorflow.keras.applications import vgg19
from tensorflow.keras.preprocessing import image
from tensorflow.keras.applications.vgg19 import preprocess_input
import faiss
import cv2


def query_yes_no(question, default="yes"):
    """Ask a yes/no question via raw_input() and return their answer.

    "question" is a string that is presented to the user.
    "default" is the presumed answer if the user just hits <Enter>.
        It must be "yes" (the default), "no" or None (meaning
        an answer is required of the user).

    The "answer" return value is True for "yes" or False for "no".
    """
    valid = {"yes": True, "y": True, "ye": True,
             "no": False, "n": False}
    if default is None:
        prompt = " [y/n] "
    elif default == "yes":
        prompt = " [Y/n] "
    elif default == "no":
        prompt = " [y/N] "
    else:
        raise ValueError("invalid default answer: '%s'" % default)

    while True:
        sys.stdout.write(question + prompt)
        choice = input().lower()
        if default is not None and choice == '':
            return valid[default]
        elif choice in valid:
            return valid[choice]
        else:
            sys.stdout.write("Please respond with 'yes' or 'no' "
                             "(or 'y' or 'n').\n")


model = vgg19.VGG19(weights="imagenet", include_top=False, pooling="avg")


def extract_features(img_path, model):
    img = image.load_img(img_path, target_size=(224, 224))
    img = image.img_to_array(img)
    img = np.expand_dims(img, axis=0)
    img = preprocess_input(img)
    feature = model.predict(img)
    return feature.flatten()


image_dir = "images/"
image_paths = [
    os.path.join(image_dir, f) for f in os.listdir(image_dir) if f.endswith(".jpg")
]
features = []

if os.path.exists("image_index.bin"):
    if query_yes_no("Load the index?", default="yes"):
        index = faiss.read_index("image_index.bin")
    else:
        for image_path in image_paths:
            img_feature = extract_features(image_path, model)
            features.append(img_feature)

        features = np.array(features)

        d = features.shape[1]
        index = faiss.IndexFlatL2(d)
        index.add(features)

        if query_yes_no("Save the index?", default="yes"):
            faiss.write_index(index, "image_index.bin")
else:
    for image_path in image_paths:
        img_feature = extract_features(image_path, model)
        features.append(img_feature)

    features = np.array(features)

    d = features.shape[1]
    index = faiss.IndexFlatL2(d)
    index.add(features)

    if query_yes_no("Save the index?", default="yes"):
        faiss.write_index(index, "image_index.bin")


def find_similar_images(query_image_path, index, k=6):
    query_feature = extract_features(query_image_path, model).reshape(1, -1)
    distance, indices = index.search(query_feature, k)
    return distance.flatten(), indices.flatten()


query_image_path = "query_image.jpg"
similar_image_distance, similar_image_indices = find_similar_images(query_image_path, index)

# display the results
for i, idx in enumerate(similar_image_indices):
    similar_image_path = image_paths[idx]
    img = cv2.imread(similar_image_path)
    similarity = 1 - similar_image_distance[i] / np.max(similar_image_distance)
    cv2.imshow(f"Similarity: {similarity * 100:.2f}% ({i+1}/{len(similar_image_indices)})", img)
    cv2.waitKey(0)
    cv2.destroyAllWindows()